================
@@ -5,15 +5,16 @@ define i32 @select_bit(ptr addrspace(1) %p, i32 %g, i32 %acc)
{
; CHECK-LABEL: define i32 @select_bit(
; CHECK-SAME: ptr addrspace(1) [[P:%.*]], i32 [[G:%.*]], i32 [[ACC:%.*]])
#[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[ENTRY:.*:]]
-; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[G]], i64 0
-; CHECK-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[TMP0]], <2 x i32>
poison, <2 x i32> zeroinitializer
-; CHECK-NEXT: [[TMP2:%.*]] = and <2 x i32> [[TMP1]], <i32 1, i32 2>
-; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr addrspace(1) [[P]], align 4
-; CHECK-NEXT: [[TMP4:%.*]] = icmp eq <2 x i32> [[TMP2]], zeroinitializer
-; CHECK-NEXT: [[TMP5:%.*]] = select <2 x i1> [[TMP4]], <2 x i32>
zeroinitializer, <2 x i32> [[TMP3]]
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i32> [[TMP5]], i64 0
+; CHECK-NEXT: [[P1:%.*]] = getelementptr inbounds i8, ptr addrspace(1)
[[P]], i64 4
----------------
jayfoad wrote:
Can you explain why this non-vectorized version is better for AMDGPU? Maybe
there is a way to get equally good codegen from the vectorized version?
https://github.com/llvm/llvm-project/pull/232877
_______________________________________________
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits